Reinforced variable selection

• Published in: Statistical theory and related fields pp. 1 - 18
• Main Authors: Le, Yuan, Bai, Yang, Zhou, Fan
• Format: Journal Article
• Language: English
• Published: Taylor & Francis Group 20.06.2025
• Subjects: natural policy gradient; reinforcement learning; Variable selection
• ISSN: 2475-4269; 2475-4277
• DOI: 10.1080/24754269.2025.2516346

Variable selection identifies the best subset of covariates when building the prediction model, among all possible subsets. In this paper, we propose a novel reinforced variable selection method, called ‘Actor-Critic-Predictor’. The actor takes an action to choose variables and the predictor evaluates the action based on a well-designed reward function, where the critic learns the reward baseline. We model the variable selection process as a multi-armed bandit and update the subset of selected variables using a natural policy gradient algorithm. We provide an analytical framework on how different errors impact the performance of our method theoretically. Large amounts of experiments on synthetic and real datasets show that the proposed framework is easily implemented and outperforms classical variable selection methods in a wide range of scenarios.